可拓AdaBoost算法对预测结果的改进  

Extenics Ada Boost for modifying the prediction of classification algorithm

作　　者：朱弘扬 高红[1] 刘巍[1] 丁怡[2] 

机构地区：[1]大连海事大学数学系,辽宁大连116026 [2]广东海洋大学工程学院,广东湛江524088

出　　处：《辽宁工程技术大学学报（自然科学版）》2016年第9期993-997,共5页Journal of Liaoning Technical University (Natural Science)

基　　金：辽宁省自然科学基金项目(2015020033)

摘　　要：为解决传统分类算法的预测结果所提供的信息不够丰富的问题,采用理论分析和实验相结合的方法,融合Ada Boost算法和可拓学理论,提出一种全新的以关联函数的形式对分类结果进行表达的可拓Ada Boost算法.以Ada Boost算法为框架,结合可拓理论中可拓模型和关联函数的理论,寻找物元的本质特征元并以此作为弱分类器,提出了根据事物隶属所有类别的关联度来判断事物类别的方法.实验表明,可拓Ada Boost算法不仅用关联函数优化了传统分类算法预测结果为关于某一类别"是"或"不是"的简单表述,而且在适用问题中可以使分类结果更准确.To resolve the question that traditional classification algorithm can＇t provide a result with abundant information, this paper proposed the Extenics AdaBoost algorithm which expresses the results through dependent function by using the method of theoretical analysis and experimental surveys. Combining the extension model and dependent function, this study is searching for the most possible essential characteristic as the weak classification algorithm and classifies things with the dependent function based on the frame of AdaBoost. The prediction of classical classification algorithm about certain category, ＂Yes＂ or ＂No＂, is replaced by the dependent function of Extenics AdaBoost with a prediction, which makes the result of classification more accurate.

关 键 词：ADA Boost 弱学习算法 可拓学 关联函数 物元 

分 类 号：U248[交通运输工程—道路与铁道工程]